A minimax robust learning framework for uncertainty-aware indoor localization in LoRaWAN networks using adaptive signal filtering and environmental intelligence

Authors

  • S. Daphine Sherin Mancy Author
  • Dr.D. David Neels Ponkumar Author
  • Dr.G. Aravind Swaminathan Author
  • Dr.K. Balamurugan Author

Keywords:

Minimax estimation, robust filtering, indoor localization, LoRaWAN, adaptive filtering, uncertainty quantification.

Abstract

LoRaWAN is increasingly used for indoor Internet-of-Things sensing, but its received signal strength indicator (RSSI) is heavily perturbed by multipath, human blockage, wall structure, and microclimate dynamics, so ranging methods built only to minimize average error give no guarantee about the worst case. This paper treats indoor LoRaWAN ranging as a minimax robust estimation problem. We couple an environment-aware multi-wall path-loss model with a Minimax Robust Adaptive Filter (MRAF), which applies a Huber-type bounded-influence correction to the RSSI innovation together with an environmentally driven process-noise schedule, and we prove that the resulting one-step innovation risk stays bounded over an epsilon-contamination neighborhood of the nominal noise model, unlike the corresponding unbounded Kalman correction. The framework is tested on 1,715,869 aggregated uplinks collected over roughly eight months from six statically deployed end devices, 8 to 40 meters from a single gateway, with 0 to 2 concrete walls and 0 to 5 wooden walls per link. Against raw RSSI, a moving-average filter, and a standard Kalman filter, the proposed method reduces mean ranging RMSE from 92.0 m to 8.4 m and reduces the worst-case RMSE across the six links from 160.9 m to 11.7 m, degrading by only 15.8% under adverse channel conditions versus 31.5% for the standard Kalman filter. These results indicate that explicit minimax-robust, environment-aware filtering is a theoretically grounded and practically inexpensive route to uncertainty-aware indoor localization.

Downloads

Download data is not yet available.

Downloads

Published

2026-09-22

How to Cite

A minimax robust learning framework for uncertainty-aware indoor localization in LoRaWAN networks using adaptive signal filtering and environmental intelligence. (2026). Minimax Theory and Its Applications, 11(2s), 447-460. https://journalmta.com/index.php/jmta/article/view/427